阅读说明：本技术 提高室内相邻网格定位准确率的方法、装置、设备和介质 (Method, device, equipment and medium for improving positioning accuracy of indoor adjacent grids ) 是由 潘维蔚 于 2019-03-15 设计创作，主要内容包括：本申请提供的一种提高室内相邻网格定位准确率的方法、装置、设备和介质,通过对相邻网格进行数据采集；根据无线保真RSSI数据的自身特征进行去除异常值；利用线性判别算法(LDA)在低维情况下进行排列组合,对所得的若干个概率值求和；通过门限的设置,对在线定位阶段的新数据进行了额外的约束。本申请能够去除异常值,提供全新的一种低维组合的方法对LDA进行处理,提高了相邻网格的区分准确率。(According to the method, the device, the equipment and the medium for improving the positioning accuracy of the indoor adjacent grids, data acquisition is carried out on the adjacent grids; removing abnormal values according to the characteristics of the wireless fidelity RSSI data; utilizing a Linear Discriminant Algorithm (LDA) to perform permutation and combination under the condition of low dimension, and summing a plurality of obtained probability values; by setting the threshold, additional constraint is carried out on the new data in the online positioning stage. The method and the device can remove abnormal values, provide a brand-new low-dimensional combination method for processing the LDA, and improve the distinguishing accuracy of the adjacent grids.)

1. A method for improving indoor neighboring grid positioning accuracy, the method comprising:

selecting a pair of adjacent grids and acquiring a training data set formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time;

normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing;

performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold;

and acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

2. The method of claim 1, wherein the step of substituting each k-dimensional RSSI value into the normalized training data set for outlier processing comprises:

searching N points with the nearest distance from the test point corresponding to each k-dimensional RSSI value by using the Euclidean distance;

judging whether the proportion of points, corresponding to the grids of the N points, which are consistent with the grids corresponding to the test points reaches a threshold value or not; if not, judging that the k-dimensional RSSI value corresponding to the test point is abnormal for removal.

3. The method of claim 2, wherein the method for finding the N nearest points to the test point corresponding to each k-dimensional RSSI value further comprises: mahalanobis distance, manhattan distance, hamming distance, and the like.

4. The method as claimed in claim 1, wherein the method for reducing the k-dimensional RSSI values into k sets of k-1-dimensional RSSI values and performing LDA operation to obtain k probability values, and determining the grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold comprises:

presetting a constraint threshold as lambda, wherein the value range of the lambda is (0.5, 1);

if the sum of the k probability values is larger than or equal to k x lambda, judging that the grid corresponding to the k-dimensional RSSI value is a first grid;

if the sum of the k probability values is less than k (1-lambda), judging that the grid corresponding to the k-dimensional RSSI value is a second grid;

and if the sum of the k probability values is less than k x lambda and is more than or equal to k x (1-lambda), directly judging the grid corresponding to each k-dimensional RSSI value according to the LDA operation.

5. The method of claim 4, wherein the probability value is different according to different test scenarios; the constraint threshold can be adjusted according to different scenes.

6. The method of claim 1, wherein the pair of adjacent grids can be grids having edge lengths in a range of 1-2 meters.

7. The method of claim 1, wherein the k-dimensional RSSI values generated for the same or similar time are calibrated simultaneously.

8. The method of claim 1, wherein the dimension reduction of the k-dimensional RSSI values into k sets of k-1-dimensional RSSI values is performed according to a certain sequence and cannot be changed at will; and each k-1 dimensional RSSI value corresponds to different AP access points respectively.

9. The method of claim 1, wherein the pair of adjacent grids is two grids immediately adjacent according to a division in a fingerprint positioning algorithm.

10. An apparatus for improving indoor adjacent grid positioning accuracy, the apparatus comprising:

the training module is used for selecting a pair of adjacent grids and acquiring a training data set formed by a plurality of groups of k-dimensional RSSI values which are matched in the adjacent grids and generated by k AP access points according to different time; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold;

and the test module is used for acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

11. The apparatus of claim 1, wherein the function of substituting each k-dimensional RSSI value into the normalized training data set for outlier processing comprises:

searching N points with the nearest distance from the test point corresponding to each k-dimensional RSSI value by using the Euclidean distance;

judging whether the proportion of points, corresponding to the grids of the N points, which are consistent with the grids corresponding to the test points reaches a threshold value or not; if not, judging that the k-dimensional RSSI value corresponding to the test point is abnormal for removal.

12. The apparatus of claim 11, wherein the method for finding the N nearest points to the test point corresponding to each k-dimensional RSSI value further comprises: any one of mahalanobis distance, manhattan distance, and hamming distance.

13. The apparatus of claim 1, wherein the function of reducing the k-dimensional RSSI values into k sets of k-1-dimensional RSSI values and performing LDA operation to obtain k probability values, and determining the grid corresponding to each k-dimensional RSSI value according to a preset constraint threshold comprises:

presetting a constraint threshold as lambda, wherein the value range of the lambda is (0.5, 1);

if the sum of the k probability values is larger than or equal to k x lambda, judging that the grid corresponding to the k-dimensional RSSI value is a first grid;

if the sum of the k probability values is less than k (1-lambda), judging that the grid corresponding to the k-dimensional RSSI value is a second grid;

and if the sum of the k probability values is less than k x lambda and is more than or equal to k x (1-lambda), directly judging the grid corresponding to each k-dimensional RSSI value according to the LDA operation.

14. The apparatus of claim 13, wherein the probability value is different according to different test scenarios; the constraint threshold can be adjusted according to different scenes.

15. The apparatus of claim 1, wherein the pair of adjacent grids can be grids having a side line length in a range of 1-2 meters.

16. The apparatus of claim 1, wherein the k-dimensional RSSI values generated for the same or similar time are calibrated simultaneously.

17. The apparatus of claim 1, wherein the dimension reduction of k-dimensional RSSI values into k sets of k-1-dimensional RSSI values is performed according to a certain sequence and cannot be changed at will; and each k-1 dimensional RSSI value corresponds to different AP access points respectively.

18. The apparatus of claim 1, wherein the pair of adjacent grids is two grids immediately adjacent to each other according to a division in a fingerprint positioning algorithm.

19. An apparatus for improving indoor adjacent grid positioning accuracy, the apparatus comprising: a memory, and a processor;

the memory is used for storing a computer program, and the processor is used for executing the computer program stored by the memory, so that when being executed, the device realizes the method for improving the indoor adjacent grid positioning accuracy according to any one of claims 1 to 9.

20. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of improving the accuracy of the positioning of indoor neighboring cells according to any one of claims 1 to 9.

Technical Field

The present application relates to the field of wireless communication network technology. And more particularly, to a method, apparatus, device, and medium for improving the accuracy of indoor neighboring grid positioning.

Background

A Wireless-Fidelity (WiFi) system refers to a technology that can connect terminals such as notebooks and mobile devices to each other in a Wireless manner through some fixed Access Points (APs). Usually, APs are deployed at some convenient indoor location, and the locations of the APs are known by a system or network administrator, so that WiFi-connected devices can communicate directly or indirectly through the APs.

The WiFi positioning technology refers to that WiFi simultaneously realizes a positioning function in addition to a communication function. WiFi is widely used in various large or small buildings such as homes, hotels, cafes, airports, shopping malls, etc., which makes WiFi a most attractive wireless technology in the field of indoor positioning.

Rssi (received Signal Strength indication), which is an optional part of the radio transmission layer, is used to determine the link quality and whether to increase the broadcast transmission Strength. RSSI is required for most wireless communication devices to operate properly, and many communication systems require RSSI information for sensing the quality of a link, performing handover, adapting transmission rate, and the like. Almost all commercial wireless devices, including smart phones, wireless sensors, rfidreaders, bluetooth, LTE, etc., support RSSI data acquisition. The acquisition of RSSI of WiFi signals is simple, and the RSSI depends on the location of the receiver. RSSI is not affected by signal bandwidth and does not require high bandwidth, so RSSI is a very popular signal feature and is widely used in positioning.

The fingerprint positioning algorithm is a set of algorithms provided based on different signal intensity information formed at different positions and formed by signal reflection and refraction in indoor complex environment. The fingerprint positioning algorithm is divided into an off-line stage and an on-line stage. In the off-line training stage, an indoor area is divided into grids (the distance is 1 to 2m), grid sampling points are sampled one by using receiving equipment, each grid corresponds to a unique fingerprint, the fingerprint can be single-dimensional or multi-dimensional, is used for receiving one feature or a plurality of features of information or signals, can record the position of the point, the obtained RSSI (received signal strength indicator), AP (access point) address and the like generally, and is the most common feature of the RSSI, fingerprint data are processed (filtered, averaged and the like), and a fine-grained fingerprint database is established at a large number of known positions. In the online positioning stage, a user holds the mobile equipment to move in a positioning area, acquires the current RSSI and the AP address in real time, and uploads the information to a server for matching.

However, both the traditional method and the method using machine learning have the problem that the accumulated positioning accuracy is too low in the error range lower than 2m, and the effect of distinguishing adjacent grids is poor.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, it is an object of the present application to provide a method, an apparatus, a device and a medium for improving the positioning accuracy of indoor neighboring cells, so as to solve the problem of poor accuracy of distinguishing neighboring cells in the existing WiFi positioning system.

To achieve the above and other related objects, the present application provides a method for improving the accuracy of positioning an indoor neighboring grid, comprising: selecting a pair of adjacent grids and acquiring a training data set formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold; and acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

In an embodiment of the present application, the method for performing outlier processing by substituting each k-dimensional RSSI value into the normalized training data set includes: searching N points with the nearest distance from the test point corresponding to each k-dimensional RSSI value by using the Euclidean distance; judging whether the proportion of points, corresponding to the grids of the N points, which are consistent with the grids corresponding to the test points reaches a threshold value or not; if not, judging that the k-dimensional RSSI value corresponding to the test point is abnormal for removal.

In an embodiment of the present application, the method for finding N points closest to the test point corresponding to each k-dimensional RSSI value further includes: any one of mahalanobis distance, manhattan distance, and hamming distance.

In an embodiment of the present application, the method for reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values and performing LDA operation to obtain k probability values, and determining the grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold includes: presetting a constraint threshold as lambda, wherein the value range of the lambda is (0.5, 1); if the sum of the k probability values is larger than or equal to k x lambda, judging that the grid corresponding to the k-dimensional RSSI value is a first grid; if the sum of the k probability values is less than k (1-lambda), judging that the grid corresponding to the k-dimensional RSSI value is a second grid; and if the sum of the k probability values is less than k x lambda and is more than or equal to k x (1-lambda), directly judging the grid corresponding to each k-dimensional RSSI value according to the LDA operation.

In an embodiment of the present application, the probability value is different according to different test scenarios; the constraint threshold can be adjusted according to different scenes.

In an embodiment of the present application, the pair of adjacent grids can be grids having a border length in a range of 1 meter to 2 meters.

In an embodiment of the present application, a plurality of k-dimensional RSSI values generated at the same time or at an approximate time are calibrated simultaneously.

In an embodiment of the present application, the dimension reduction of each k-dimensional RSSI value into k sets of k-1-dimensional RSSI values is performed according to a certain sequence, and cannot be changed at will; and each k-1 dimensional RSSI value corresponds to different AP access points respectively.

In an embodiment of the present application, the pair of adjacent grids is based on two grids immediately adjacent to each other divided in the fingerprint location algorithm.

To achieve the above and other related objects, the present application provides an apparatus for improving the accuracy of positioning of indoor neighboring cells, the apparatus comprising: the training module is used for selecting a pair of adjacent grids and acquiring a training data set formed by a plurality of groups of k-dimensional RSSI values which are matched in the adjacent grids and generated by k AP access points according to different time; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold; and the test module is used for acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

In an embodiment of the application, the function of substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing includes: searching N points with the nearest distance from the test point corresponding to each k-dimensional RSSI value by using the Euclidean distance; judging whether the proportion of points, corresponding to the grids of the N points, which are consistent with the grids corresponding to the test points reaches a threshold value or not; if not, judging that the k-dimensional RSSI value corresponding to the test point is abnormal for removal.

In an embodiment of the present application, the method for finding N points closest to the test point corresponding to each k-dimensional RSSI value further includes: any one of mahalanobis distance, manhattan distance, and hamming distance.

In an embodiment of the present application, the function of reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values and performing LDA operation to obtain k probability values, and determining the grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold includes: presetting a constraint threshold as lambda, wherein the value range of the lambda is (0.5, 1); if the sum of the k probability values is larger than or equal to k x lambda, judging that the grid corresponding to the k-dimensional RSSI value is a first grid; if the sum of the k probability values is less than k (1-lambda), judging that the grid corresponding to the k-dimensional RSSI value is a second grid; and if the sum of the k probability values is less than k x lambda and is more than or equal to k x (1-lambda), directly judging the grid corresponding to each k-dimensional RSSI value according to the LDA operation.

In an embodiment of the present application, the probability value is different according to different test scenarios; the constraint threshold can be adjusted according to different scenes.

In an embodiment of the present application, the pair of adjacent grids can be grids having a border length in a range of 1 meter to 2 meters.

In an embodiment of the present application, a plurality of k-dimensional RSSI values generated at the same time or at an approximate time are calibrated simultaneously.

In an embodiment of the present application, the dimension reduction of each k-dimensional RSSI value into k sets of k-1-dimensional RSSI values is performed according to a certain sequence, and cannot be changed at will; and each k-1 dimensional RSSI value corresponds to different AP access points respectively.

In an embodiment of the present application, the pair of adjacent grids is based on two grids immediately adjacent to each other divided in the fingerprint location algorithm.

To achieve the above and other related objects, the present application provides an apparatus for improving indoor adjacent grid positioning accuracy, the apparatus comprising: a memory, and a processor; the memory is used for storing a computer program, and the processor is used for executing the computer program stored by the memory, so that the device can realize the method for improving the indoor adjacent grid positioning accuracy when being executed.

To achieve the above objects and other related objects, the present application provides a computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements the method for improving the accuracy of indoor neighboring grid positioning as described above.

As described above, according to the method, the apparatus, the device, and the medium for improving the positioning accuracy of the indoor adjacent grids, a pair of adjacent grids is selected and a training data set composed of a plurality of k-dimensional RSSI values generated by k AP access points according to different times is obtained; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold; and acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

Has the following beneficial effects:

abnormal values can be removed, a brand-new low-dimensional combination method is provided for processing the LDA, and the distinguishing accuracy of adjacent grids is improved.

Drawings

Fig. 1 is a flowchart illustrating a method for improving the accuracy of positioning an indoor neighboring grid according to an embodiment of the present invention.

Fig. 2 is a schematic view illustrating a fingerprint positioning scenario in an embodiment of the present application.

Fig. 3 is a block diagram illustrating an apparatus for improving the accuracy of positioning an indoor neighboring grid according to an embodiment of the present invention.

Fig. 4 is a schematic structural diagram illustrating an apparatus for improving the accuracy of positioning an indoor neighboring grid according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present application is provided by way of specific examples, and other advantages and effects of the present application will be readily apparent to those skilled in the art from the disclosure herein. The present application is capable of other and different embodiments and its several details are capable of modifications and/or changes in various respects, all without departing from the spirit of the present application. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.

It is noted that in the following description, reference is made to the accompanying drawings which illustrate several embodiments of the present application. It is to be understood that other embodiments may be utilized and that mechanical, structural, electrical, and operational changes may be made without departing from the spirit and scope of the present application. The following detailed description is not to be taken in a limiting sense, and the scope of embodiments of the present application is defined only by the claims of the issued patent. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. Spatially relative terms, such as "upper," "lower," "left," "right," "lower," "below," "lower," "above," "upper," and the like, may be used herein to facilitate describing one element or feature's relationship to another element or feature as illustrated in the figures.

In this application, unless expressly stated or limited otherwise, the terms "mounted," "connected," "secured," "retained," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the present application can be understood by those of ordinary skill in the art as appropriate.

Also, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms "comprises," "comprising," and/or "comprising," when used in this specification, specify the presence of stated features, operations, elements, components, items, species, and/or groups, but do not preclude the presence, or addition of one or more other features, operations, elements, components, items, species, and/or groups thereof. The terms "or" and/or "as used herein are to be construed as inclusive or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: a; b; c; a and B; a and C; b and C; A. b and C ". An exception to this definition will occur only when a combination of elements, functions or operations are inherently mutually exclusive in some way.

In view of the interference problem existing in the WSN network, the present application provides a method, an apparatus, a device and a medium for improving the positioning accuracy of indoor adjacent grids, by performing data acquisition on the adjacent grids; removing abnormal values according to the characteristics of the wireless fidelity RSSI data; utilizing a Linear Discriminant Algorithm (LDA) to perform permutation and combination under the condition of low dimension, and summing a plurality of obtained probability values; through the setting of the threshold, additional constraint is carried out on new data in the online positioning stage, and therefore the problem that the accuracy of distinguishing adjacent grids by a WiFi positioning system is poor is solved.

Fig. 1 is a schematic flow chart illustrating a method for improving the accuracy of positioning an indoor neighboring grid according to an embodiment of the present invention. As shown, the method comprises:

step S101: and selecting a pair of adjacent grids and acquiring a training data set formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time matched in the grids.

For example, as shown in fig. 2, as a fingerprint database built from a fingerprint location algorithm. As shown, AP is WiFi, and RP is a reference point selected for offline acquisition. The position fingerprint method in indoor positioning measures signal characteristics (signal strength of Wi-Fi) of each position in advance once and stores the signal characteristics into a fingerprint database. When positioning, the current signal characteristics are matched with those in the fingerprint database, so as to determine the position.

For example, in a fingerprint database with a grid side length of 3m, the terminal location of the indoor position to be measured is more accurate, but when the grid side length is reduced to 2m or 1m, the location accuracy is greatly reduced.

Further, when the length of the grid edge is to be accurate from 3m to 2m or 1m, the terminal at the position to be measured in the grid may only be able to determine in several grids, and may not be able to accurately determine in which smaller grid, so that the method for improving the positioning accuracy of the indoor adjacent grid is proposed in the present application to solve the problem.

In view of the above process, a pair of adjacent meshes is first selected, thereby further determining in which mesh in particular. It should be noted that the terminal at the original position to be measured should be located in a grid (e.g. a grid with a side length of 3 m) with a larger side length formed by adjacent grids (e.g. a grid with a side length of 2 m).

In an embodiment of the present application, the pair of adjacent grids is based on two grids immediately adjacent to each other divided in the fingerprint location algorithm.

In an embodiment of the present application, the pair of adjacent grids can be grids having a border length in a range of 1 meter to 2 meters.

As mentioned above, the AP access point is a wireless access point, which may be WiFi, for example, and is usually deployed in some convenient indoor location, and the location of the AP is known by a system or a network administrator, and devices connected to WiFi can communicate directly or indirectly through the AP.

Rssi (received Signal Strength indication), which is an optional part of the radio transmission layer, is used to determine the link quality and whether to increase the broadcast transmission Strength. RSSI is required for most wireless communication devices to operate properly, and many communication systems require RSSI information for sensing the quality of a link, performing handover, adapting transmission rate, and the like. And almost all commercial wireless devices, including smart phones, wireless sensors, rfidreaders, bluetooth, LTE, etc., support the collection of RSS data. The acquisition of RSSI of WiFi signals is simple, and the RSSI depends on the location of the receiver. RSSI is not affected by signal bandwidth and does not require high bandwidth, so RSSI is a very popular signal feature and is widely used in positioning.

In this embodiment, through wireless interconnection between the k AP access points and the terminal at the location to be measured, a plurality of sets of k-dimensional RSSI signal strength values, for example [100, 150, 30, 80], corresponding to the terminal at the location to be measured can be obtained.

In an embodiment of the present application, a plurality of k-dimensional RSSI values generated at the same time or at an approximate time are calibrated simultaneously.

Here, multiple sets of k-dimensional RSSI values generated for the same time or approximate time need to be calibrated so that k APs each receive one RSSI value at a time.

Step S102: and carrying out normalization processing on the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to carry out abnormal value processing.

In this embodiment, the training data set is normalized for easy operation.

In an embodiment of the present application, the method for performing outlier processing by substituting each k-dimensional RSSI value into the normalized training data set includes:

A. and searching N points which are closest to the test point corresponding to each k-dimensional RSSI value by using the Euclidean distance.

In an embodiment of the present application, the euclidean distance may be replaced by any one of a mahalanobis distance, a manhattan distance, and a hamming distance to find N points closest to the test point corresponding to each k-dimensional RSSI value.

B. Judging whether the proportion of points, corresponding to the grids of the N points, which are consistent with the grids corresponding to the test points reaches a threshold value or not; if not, judging that the k-dimensional RSSI value corresponding to the test point is abnormal for removal.

In this embodiment, the threshold may be one-half, one-third, two-thirds, or the like.

For example, 50 points closest to the test point corresponding to each k-dimensional RSSI value are found, and if the number of points consistent with the grid corresponding to the test point is less than 25 (one half), it is determined that the k-dimensional RSSI value corresponding to the test point is abnormal, that is, the group of data needs to be removed and cannot be subjected to the required operation.

The abnormal value processing of the data set is a kind of filtering of the data set, because various interferences, such as human walking, environmental signal noise, etc., can be received in the actual scene, therefore, the accuracy of the data set to be operated can be ensured by performing the abnormal value processing, and the accuracy of the subsequent positioning operation and judgment can be greatly improved.

Step S103: performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold.

In an embodiment, this step may actually perform two processes simultaneously: 1) performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; 2) and reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold.

Linear Discriminant Analysis (LDA) is a machine learning algorithm, which projects a high-dimensional pattern (for example, RSSI samples of multiple APs, different APs corresponding to different dimensions) to an optimal discriminative vector space to achieve the effect of extracting classification information and compressing feature space dimensions, so that an inter-class scattering matrix of the projected pattern samples is maximized and an intra-class scattering matrix is minimized at the same time, that is, a minimum intra-class distance and a maximum inter-class distance in a new space after projection are ensured, that is, high-dimensional data of adjacent grids has optimal separability in the space. Machine learning is a method that can give the machine the ability to learn and thus make it perform functions that cannot be done by direct programming. In a practical sense, machine learning is a method of training a model by using data and then using the model to predict.

The specific method for the processing mode of 2) is as follows:

presetting a constraint threshold as lambda, wherein the value range of the lambda is (0.5, 1);

if the sum of the k probability values is larger than or equal to k x lambda, judging that the grid corresponding to the k-dimensional RSSI value is a first grid;

if the sum of the k probability values is less than k (1-lambda), judging that the grid corresponding to the k-dimensional RSSI value is a second grid;

and if the sum of the k probability values is less than k x lambda and is more than or equal to k x (1-lambda), directly judging the grid corresponding to each k-dimensional RSSI value according to the LDA operation.

In this embodiment, the k-dimensional RSSI values are reduced to k sets of k-1-dimensional RSSI values, specifically, one AP (one-dimensional) is removed from each set in sequence to form k-1-dimensional data.

For example, [10,7,8,9,6] becomes [10,7,8,9], [10,7,8,6], [10,7,9,6], [10,8,9,6], [7,8,9,6 ].

In an embodiment of the present application, the dimension reduction of each k-dimensional RSSI value into k sets of k-1-dimensional RSSI values is performed according to a certain sequence, and cannot be changed at will; and each k-1 dimensional RSSI value corresponds to different AP access points respectively.

For example, the first grid is set to tag 0 and the 2 nd grid is set to tag 1.

The training is that since there are labels of 0 and 1, and if there is this label, we can obtain five LDAs with different internal parameters, and this parameter difference means that the projected mapping vectors are different, these five data sets are very different data sets, and different data sets have completely different projection vectors in the LDA algorithm, so there are 5 different low-dimensional LDAs, then if there is new test 5-dimensional data [100,400,500,200,300], we do not know whether its label is 0 or 1, then we also divide the data into 5 groups of 4-dimensional data, and the order cannot be disordered, and test these five separately with LDAs of 5 different mapping vectors already obtained, this time we obtain probabilities of 0 or 1 generated inside 5 different LDAs, and we add these probabilities to see if we do not satisfy our threshold, and then use threshold decision. In fact, most data satisfy the threshold described in the present application, and for a few that do not satisfy both thresholds, the initial first path may be used to directly determine if LDA is 0 or 1.

In this embodiment, an LDA algorithm is performed on k sets of k-1 dimensional data to obtain probabilities P with predicted values of 0 output by the actual application algorithm₁、P₂…、P_kThen the probability corresponding to the predicted value of 1 is (1-P)₁)、(1-P₂)…(1-P_k) The summation is completed.

Note that the variance of the signal intensity values due to the change in spatial position is almost the same in the respective grids. The variance of adjacent grid data is mainly divided into two parts of space position and self energy fluctuation, the self energy fluctuation is mainly determined by energy change of an emission point and arrival time change, therefore, the self energy fluctuation of an adjacent grid can be considered to be approximate to that of the same AP emission point, the one-dimensional data variance is approximate, the multi-dimensional data variance is simple algebraic addition of the one-dimensional data variance due to independence, the variance of the multi-dimensional data in each grid is approximate, and the LDA algorithm is used for highlighting that the inter-class distance is reasonable for processing the adjacent grid with the center distance being less than 2 m.

Setting a constraint threshold: entering a threshold comparison unit, if the point satisfies the following formula, taking the M judgment value as 0:

P₁+P₂+…+P₄≧k*λ；

correspondingly, if the following equation is satisfied, the M decision value takes 1:

P₁+P₂+…+P₄≦k*(1-λ)；

if the probability summation does not satisfy the two expression ranges, the initial LDA predicted value is considered to be right. Wherein, the value range of the lambda is (0.5, 1). For example, λ is 0.8.

In an embodiment of the present application, the probability value is different according to different test scenarios; the constraint threshold can be adjusted according to different scenes.

It should be noted that the probability value is a parameter for calculating the probability value through training of the training set, so that the corresponding probability value can be obtained by substituting the measured data set.

Step S104: and acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

And substituting the acquired measured data set into the step S103 to determine the grid corresponding to each k-dimensional RSSI value.

Of course, the method can be extended to other outdoor scenes as long as two grids meet the requirement of a fixed small area.

According to the method, threshold constraint of low-dimensional combination of the removed abnormal values is completed, the RSSI data of the new multi-dimensional AP can be almost completely and accurately distinguished on which side of the adjacent grid is located under the same environment, the accuracy rate reaches about 99%, the RSSI data is obviously improved by more than 5% compared with other algorithms, and the RSSI data of the new multi-dimensional AP under different environments is also greatly improved. Finally, the method does not increase the receiving end overhead and improves the binary distinguishing accuracy of adjacent grids.

The following is further illustrated by the specific examples.

The adjacent grids are selected to be adjacent parts inside and outside an office door of a building with a plurality of office facilities and personnel, a Huashi mobile phone NEM-10 is selected as a receiving device in the experiment, a WiFi module is arranged in the mobile phone, 4 APs are placed in the office, and the AP is respectively connected with 4 personal computers through HL-340USB serial ports. Before testing, program is written into the AP kernel and is fully debugged, so that when the AP receives a reverse response signal of the mobile phone, RSS information, the MAC address of the mobile phone and the moment of receiving the response signal are output to a personal computer through a serial port, and simultaneous calibration is carried out. The personal computer keeps records and integrates the information of 4 personal computers during the test.

A. The real environment used in the experiment is an office environment, wherein an indoor grid of the office and an outdoor grid of a corridor are selected as adjacent grid points, the selected adjacent grid points are an inner grid taking the door as a boundary line in the door and an outer grid corresponding to the outside of the door (the grid is a square with 1m1m, namely the central distance of the adjacent grids can fully ensure that the data processed by the algorithm is less than a 2m error range, the accuracy improvement when the grid discrimination is minimum is realized, meanwhile, in order to ensure that the accuracy improvement is more general, no door is blocked outside the door in the door, so that the difference of the signal intensity of the adjacent grids is smaller, the discrimination difficulty between the grids is larger), the time of carrying the mobile phone by the laboratory staff on the grids inside and outside the door is strictly equal during the experiment, the laboratory staff rotate the grids every 6 minutes (so that the balance of training data can be ensured), the experiment of different real office environments (such as the placement of objects in the office and the change of the staff), the first environment collects 2383 groups of 4-dimensional data, the second environment collects 535 groups of 4-dimensional data, the third environment collects 957 groups of 4-dimensional data, and the computer collects 3875 groups of 4-dimensional data to finish storage work after collection.

B. Removing abnormal values, normalizing a training data set which is generated by 4 AP and is subjected to time sequence calibration on 4-dimensional training data, performing abnormal value processing on the normalized data, namely substituting offline training multidimensional data into the offline training set, searching 50 points with the nearest Euclidean distance, directly judging by the Euclidean distance, and if any dimension data is detected to be an abnormal point, if at least 25 points in the 50 points with the nearest distance are consistent with a corresponding grid, judging that the data is not the abnormal point; otherwise, judging the abnormal point. Some defined outliers are removed.

C. Applying LDA algorithm to the data subjected to abnormal value processing to construct a model H, arranging and combining 4 APs, dividing into 4 groups, removing one AP (one dimension) in each group in sequence to form 3-dimensional data, performing LDA algorithm to the 4 groups of 3-dimensional data to respectively obtain the probability P of 0 predicted value output by the actual application algorithm₁、P₂、…、P₄The summation is done to see if it is greater than 3.2(λ is 0.8), or the probabilities corresponding to a predictor of 1 are (1-P), respectively₁)、(1-P₂)、…、(1-P₄) The summation is done to see if it is greater than 3.2.

D. Normalization by substituting test data, direct P₁、P₂、…、P₄If the sum is greater than 3.2 or less than 0.8, the sum of the ranges therebetween still applies to LDA

Finally, the accuracy rate of the invention is close to 0.99 under the same environment of an office, and the accuracy rate of the invention is improved under each environment when the complete training set is learned under different environments.

In some embodiments, the neural network-based WSN interference rejection method may be applied to a controller, for example: an ARM controller, an FPGA controller, an SoC controller, a DSP controller, or an MCU controller, etc. In some embodiments, the neural network-based WSN interference rejection method is also applicable to a computer that includes components such as a memory, a memory controller, one or more processing units (CPUs), a peripheral interface, RF circuitry, audio circuitry, a speaker, a microphone, an input/output (I/O) subsystem, a display screen, other output or control devices, and an external port; the computer includes, but is not limited to, Personal computers such as desktop computers, notebook computers, tablet computers, smart phones, smart televisions, Personal Digital Assistants (PDAs), and the like. In other embodiments, the WSN interference rejection method based on the neural network may be further applied to a server, which may be disposed on one or more physical servers according to various factors such as functions, loads, and the like, and may be formed by a distributed or centralized server cluster.

Fig. 3 is a block diagram of an apparatus for improving the accuracy of positioning an indoor neighboring grid in the embodiment of the present application. As shown, the apparatus 300 includes:

a training module 301, configured to select a pair of adjacent grids and obtain a training data set formed by multiple sets of k-dimensional RSSI values generated by k AP access points according to different times; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold;

the testing module 302 is configured to obtain a testing data set formed by multiple groups of k-dimensional RSSI values generated by k AP access points according to different times and corresponding to a terminal in the adjacent grid, and substitute the k-dimensional RSSI values in the testing data set into the previous step to determine the grid corresponding to the k-dimensional RSSI values.

It should be noted that, the embodiment of the apparatus for improving the positioning accuracy of the indoor neighboring grid of this embodiment is similar to the above embodiment of the method for improving the positioning accuracy of the indoor neighboring grid, and therefore, the detailed description is omitted.

It should be understood that the division of the modules of the above system is only a logical division, and the actual implementation may be wholly or partially integrated into one physical entity or may be physically separated. And these modules can be realized in the form of software called by processing element; or may be implemented entirely in hardware; and part of the modules can be realized in the form of calling software by the processing element, and part of the modules can be realized in the form of hardware. For example, the test module 302 may be a separately installed processing element, or may be integrated into a chip of the system, or the test module 302 may also be stored in a memory of the system in the form of program code, and a processing element of the system calls and executes the functions of the classifier model training module. Other modules are implemented similarly. In addition, all or part of the modules can be integrated together or can be independently realized. The processing element described herein may be an integrated circuit having signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more Application Specific Integrated Circuits (ASICs), or one or more microprocessors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), among others. For another example, when one of the above modules is implemented in the form of a Processing element scheduler code, the Processing element may be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

Fig. 4 is a schematic structural diagram of an apparatus for improving the accuracy of positioning an indoor neighboring grid in an embodiment of the present application. The apparatus 400 comprises: a memory 401 and a processor 402, wherein the memory 401 stores a computer program, and the processor 402 is configured to execute the computer program stored in the memory 401, so that the apparatus implements the neural network-based WSN anti-interference method as described in fig. 1 when executed.

The Memory 401 may include a Random Access Memory (RAM), and may further include a non-volatile Memory (non-volatile Memory), such as at least one disk Memory.

The Processor 402 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.

To achieve the above objects and other related objects, the present application provides a computer readable storage medium having a computer program stored thereon, which when executed by a processor, implements the method for improving the accuracy of indoor neighboring grid positioning as described in fig. 1.

The computer-readable storage medium, as will be appreciated by one of ordinary skill in the art: all or part of the steps for implementing the above method embodiments may be performed by hardware associated with a computer program. The aforementioned computer program may be stored in a computer readable storage medium. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.

In summary, a pair of adjacent grids is selected and a training data set formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time periods is obtained; normalizing the training data set, and substituting each k-dimensional RSSI value into the normalized training data set to perform abnormal value processing; performing LDA operation on the new training data set to judge grids corresponding to the k-dimensional RSSI values; and/or reducing the k-dimensional RSSI values into k groups of k-1-dimensional RSSI values, performing LDA operation to obtain k probability values, and judging grids corresponding to the k-dimensional RSSI values according to a preset constraint threshold; and acquiring a test data set which is formed by a plurality of groups of k-dimensional RSSI values generated by k AP access points according to different time and corresponds to a terminal in the adjacent grids, and substituting each k-dimensional RSSI value in the test data set into the previous step to judge the grid corresponding to each k-dimensional RSSI value.

The application effectively overcomes various defects in the prior art and has high industrial utilization value.

The above embodiments are merely illustrative of the principles and utilities of the present application and are not intended to limit the application. Any person skilled in the art can modify or change the above-described embodiments without departing from the spirit and scope of the present application. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical concepts disclosed in the present application shall be covered by the claims of the present application.